Mieux Donner

Marginal impact of donations: is your 'drop in the ocean' really useless?

Picture of Romain Barbe

Romain Barbe

Founder of Mieux Donner
Reading time : 9 minutes

In 2023, the World Bank estimated that around 700 million people were still living on less than $2.15 a day, the international extreme poverty threshold [1]. In the same year, the World Health Organisation (WHO) estimated that nearly 5 million children under the age of five were dying from largely preventable causes [2]. At the same time, the Intergovernmental Panel on Climate Change (IPCC) indicated that without a rapid reduction in emissions, it was highly likely that the +1.5°C threshold would be exceeded on a sustained basis [3].

These figures are staggering. They describe systemic, global, structural phenomena.

In this context, an individual donation of €50, €100 or €500 may seem insignificant: we spontaneously compare our contribution to the scale of the problem. However, the real impact of a donation is not limited to a subjective impression. Analyses show that, for the same amount, a donation can have up to 100 times more impact depending on the charity supported. And if every life counts, then the difference in impact should also count more.

A single donation will not, on its own, change the global curve of extreme poverty. But for the extra child who avoids a serious illness, for the family that gains security and peace of mind, the difference is considerable. Imagine a loved one who is ill: the idea that no one would help because it doesn’t make a “big difference” on a global scale would be outrageous, because it matters immensely to you. The same is true for the people receiving the aid. And this is precisely what marginal impact helps to clarify: what your donation makes possible, in addition to what would happen if there were no donation.

This tension between the scale of the problem and the actual effect of a gesture is understandable, but it is methodologically misleading: it leads to evaluating a donation on the wrong scale. To analyse it correctly, we need to introduce a central concept in economics and decision theory: marginal impact.

Rigorously define the marginal impact

Economic definition

In microeconomics, marginal analysis refers to the study of the variation in an outcome following an infinitesimal variation in a factor. Since Alfred Marshall [4], rational decision-making has been evaluated not on the basis of total quantities, but on the basis of marginal differences.

Formally, if we note:
I(D) as the total impact produced by a donation level D,
then the marginal impact corresponds to the derivative dI/dD, i.e. the change in impact generated by an additional unit of donation.

Counterfactual definition

In impact assessment, particularly in the work of MIT’s J-PAL [5], the causal effect of an intervention is defined in a counterfactual manner: it is the difference between the state of the world with intervention and the state of the world without intervention.

Applied to donations:
Marginal impact = Impact with donation – Impact without donation.

The relevant comparison is therefore not:
Is my donation proportionate to global poverty?
But:
What happens if I give, compared to if I do not give?

It is precisely this logic that we detail when we explain our selection of high-impact charities, where we set out the rigorous analysis criteria on which our recommendations are based.

Probabilistic impact: formalisation in expected utility

In many cases, the impact of a donation is not deterministic but probabilistic. A euro donated does not produce a certain and identifiable result; it modifies a probability distribution. This distinction is essential.

The theory of expected utility, formalised by John von Neumann and Oskar Morgenstern in 1944 in Theory of Games and Economic Behaviour, establishes that a rational agent faced with uncertainty must evaluate an action based on the mathematical expectation of its consequences. In other words, the rational value of an action corresponds to the sum of the possible outcomes weighted by their probability.

Formally:
Expected impact = Probability of outcome × Magnitude of outcome

This framework is now ubiquitous in public economics, finance, insurance and health economics. It allows for the comparison of interventions whose effects are uncertain but quantifiable.

Applied to giving, this means that the relevant question is not: “Does my donation produce a certain effect?” but rather: “Does my donation increase the likelihood of a socially meaningful outcome?”

Let us take a documented example. Insecticide-treated mosquito nets for malaria prevention have been the subject of rigorous meta-analyses in the Cochrane Library [11], which show a statistically significant reduction in infant mortality in the areas covered. These results are based on randomised controlled trials and the aggregation of data from a variety of contexts.

Based on this data, independent evaluators such as GiveWell [10] estimate an average cost per life saved. This estimate is not a simple mechanical division; it incorporates conservative assumptions regarding:

  • the actual rate of mosquito net usage
  • the sustainability of the effects
  • logistical losses
  • statistical uncertainty margins

Let us suppose, for illustrative purposes, that an intervention statistically saves one life for €5,000. A donation of €100 does not “save 0.02 lives”. This formulation is conceptually incorrect because human life is not divisible.

On the other hand, a donation of €100 marginally increases the aggregate probability that a life will be saved in the target population. It contributes to the statistical expectation of the collective outcome.

This logic may seem abstract, but it is standard in public policy. When a government invests in a vaccination campaign, road safety or cardiovascular disease prevention, it does not “save” a specific life. It reduces the probability of death across a population. The benefits are measured in statistical lives saved, years of life gained or QALYs (Quality-Adjusted Life Years).

Individual donations fit perfectly into this framework. They do not create visible miracles. They marginally alter a probabilistic trajectory.

At Mieux Donner, we specifically prioritise interventions with a high expected impact relative to the costs incurred, i.e. those for which an additional euro significantly increases the likelihood of a socially substantial outcome.

Additionality: a necessary condition for marginal impact

One fundamental point must be clarified: not all donations generate the same marginal impact.

In public economics, additionality refers to the portion of an effect that would not have occurred in the absence of the intervention in question. In other words, an action is said to be additional if it produces an additional result compared to the counterfactual scenario.

This concept is central to the evaluation of public policies, particularly in the analysis of subsidies, tax credits, or international financing. An intervention may appear substantial in absolute terms, but produce little additional effect if it simply replaces financing that would have existed anyway.

When applied to individual donations, additionality raises a simple but crucial question:
If I do not give, will the action still take place?

If a programme is already fully funded and an additional donation simply allows another funder to withdraw or reallocate its funds elsewhere, the actual marginal impact of the donation may be small. In this case, there is substitution rather than creation of impact.

On the other hand, if an organisation has unfunded opportunities for expansion, such as purchasing additional mosquito nets, opening a new area of intervention, or recruiting an additional health worker, then an extra euro can effectively trigger new action. Without additionality, there is no difference between the world “with a donation” and the world “without a donation”.

This analysis also involves taking diminishing returns into account. As a programme is funded, the most effective opportunities are generally exploited first. The marginal impact of an additional euro may then decrease. Rigorous evaluation consists precisely in identifying contexts where the marginal impact remains high.

Additionality is not always easily observable. It is based on analysis:

  • of the organisation’s funding structure
  • of its absorption capacity
  • the existence of projects awaiting funding
  • and real budgetary flexibility

A transparent organisation must be able to explain what an additional euro is used to fund.

Without this information, it is difficult to estimate the actual marginal impact of a donation.

Cognitive biases that distort our perception

The impression that our contribution is insignificant is not just intuitive; it is documented by cognitive psychology.

The work of Daniel Kahneman and Amos Tversky [7] has shown that our judgement is largely based on heuristics, simplifying mental rules that work well in ordinary contexts but become inadequate when dealing with extreme magnitudes. When a problem is expressed in millions or billions, our minds struggle to process the information proportionally. We spontaneously compare our actions to the phenomenon as a whole, rather than comparing them to their differential effect.

This bias can be likened to what is sometimes referred to as scale insensitivity: beyond a certain threshold, a massive quantitative increase does not generate a proportional increase in our perception of severity or usefulness. In the context of giving, this leads to a framing error: we evaluate our contribution relative to the total problem rather than relative to what it makes possible.

This scale bias not only distorts our assessment of the effectiveness of a donation; it also weakens our sense of personal responsibility. Added to this cognitive difficulty is a well-identified social mechanism: diffusion of responsibility. Experiments conducted by Darley and Latané [8] following the Kitty Genovese case [10][11] showed that the greater the number of potential witnesses, the less each individual feels personally responsible for intervening. When responsibility is perceived as collective, it tends to be diluted.

Applied to global causes such as world poverty, international health and climate change, this mechanism can reduce the sense of individual obligation: far from mobilising people, the scale of the problem can paradoxically discourage them from getting involved.

On the contrary: the example can lead to

The “bystander effect” is often summarised as follows: the more people there are, the less we act. In reality, it is more subtle than that. Yes, responsibility can be diluted… but in many situations, we also do the opposite: we align ourselves with the behaviour of others.

In other words, what matters is not only how many people could take action, but also the social signal: are others taking action, or remaining passive? When action is visible, it becomes the norm, and the norm can trigger action.

This is particularly true when it comes to giving: knowing that giving is commonplace can encourage people to give in turn, much more so than if they mistakenly believe that “almost no one” does so.

Finally, Paul Slovic’s research on “psychic numbing”, or psychological paralysis caused by numbers [9], provides further insight. It shows that our emotional response does not increase linearly with the number of victims. An identifiable life elicits a strong emotional response; a thousand abstract lives do not elicit a thousand times more emotion. As the numbers increase, our empathy tends to stabilise or even decrease.

This phenomenon creates a gap between the objective severity of a situation and our subjective perception of it. A donation that marginally increases the probability of a child avoiding a serious illness has a real effect in the world. But because this effect is statistical rather than narrative, it does not necessarily trigger a proportional emotional response.

Taken together, these cognitive and social mechanisms explain why we tend to underestimate the marginal impact of a donation. It is not that the effect is zero; it is that our psychological architecture is ill-equipped to assess large-scale probabilistic variations.

Understanding these biases does not guarantee a perfect decision. But it does at least allow us to distinguish between the impression of futility, which is psychologically explainable, and the rational assessment of impact, which is economically measurable.

Climate: structural uncertainty and long causal chains

In the climate field, analysing the marginal impact of a donation is more complex than in the case of direct interventions. The difficulty does not lie in the lack of data, but in the very structure of the problem.

The IPCC [3] establishes global physical trajectories with a high degree of scientific consensus: the relationship between cumulative emissions and warming, projections of scenarios according to different levels of mitigation, and expected impacts based on emission trajectories. Physically speaking, the link between greenhouse gas emissions and rising temperatures is well documented.

On the other hand, philanthropic levers for action rarely operate at this direct physical level. They most often involve indirect and institutional causal chains: regulatory advocacy, changes in industry standards, strategic litigation, production of expertise, transformation of economic incentives or market standards.

In other words, between a donation and an actual reduction in emissions, there are several intermediate steps:

  • Funding an organisation enables it to carry out activities (research, advocacy, coalition building, litigation).
  • This activity influences a regulatory decision or an industry standard.
  • This decision alters the economic incentives or behaviours of the actors involved.
  • These modifications result in changes in investment and, ultimately, in a reduction in emissions.

Each link in this chain has a probability of success. The final impact therefore depends on the product of these successive probabilities.

This is why climate uncertainty is said to be structural: it does not only concern the effectiveness of an isolated action, but the robustness of a sequence of interdependent events.

However, uncertainty does not imply no impact. It implies that the impact should be modelled in probabilistic rather than deterministic terms.

In economic analysis, it is common to evaluate interventions based on mathematical expectations, even when the causal chains are long. Decisions on infrastructure, research and development, and industrial policy are based on this type of reasoning: a small change in probability applied to a significant systemic outcome can generate substantial expected value.

In the case of climate change, the marginal effect of a donation can thus be understood as a change, however modest, in the probability that a more favourable regulatory or technological path will be adopted. When the global challenge – stabilising temperatures and reducing extreme risks – is enormous, small probabilistic variations can correspond to significant expected gains.

The challenge is therefore not to eliminate uncertainty, which would be impossible, but to explicitly incorporate it into reasoning. This requires:

  • explain the theory of change of the supported organisation;
  • identify critical assumptions;
  • assess the plausibility of each causal link;
  • to estimate, even approximately, the magnitude of the expected effects.

In our climate analysis, we take precisely this approach: rather than demanding unrealistic certainty or abandoning estimation altogether, we examine the consistency of the causal chain, the quality of the available evidence, and the plausibility of the mechanisms invoked.

Thinking rigorously in a context of uncertainty does not mean asserting that the impact is certain. It means determining whether the expected impact, given the probabilities and stakes, is high enough to justify the allocation of resources.

In a systemic field such as climate, rationality does not consist of seeking the most visible direct effect, but rather of identifying the levers capable of sustainably influencing global trajectories.

Methodological objections and limitations

Scientific reasoning must acknowledge its limitations.

Cost-effectiveness estimates are based on assumptions. Indirect effects can be difficult to measure. Comparisons between causes raise legitimate normative questions.

However, the alternative to an imperfect estimate is not perfection: it is the absence of analysis.

Refusing to use available data because it is uncertain amounts to substituting raw intuition for structured assessment.

Conclusion: changing the analytical framework

Comparing one’s donation to the total size of the problem is a mistake of scale. The relevant comparison is counterfactual and marginal.

If a donation increases, even slightly, the probability of a positive outcome, then it has a real impact.

In a world where public decisions are already based on probabilities, ignoring marginal logic would be inconsistent.

The question is therefore not:
“Is my donation enough to solve the problem?”
The question is:
“Does it change anything?”

When it is directed towards rigorously evaluated interventions based on solid data and demanding comparative analysis, the answer is often yes.

If you would like your contribution to be part of this approach, favouring charities with a high, documented and transparent marginal impact, you can donate via Mieux Donner and direct your donation to charities selected for the strength of their impact.

Notes and references

[1] Banque mondiale. Poverty and Inequality Platform – Extreme Poverty Data (International Poverty Line $2.15/day, 2017 PPP).
World Bank Group, données actualisées 2023.
https://www.worldbank.org/en/topic/poverty/overview

 

[2] Organisation mondiale de la santé (OMS). Global Health Observatory Data Repository – Child Mortality Estimates.
WHO, dernières estimations disponibles 2023.
https://www.who.int/data/gho

 

[3] Intergovernmental Panel on Climate Change (IPCC). Sixth Assessment Report (AR6), Synthesis Report.
IPCC, 2023.
https://www.ipcc.ch/report/ar6/syr/

 

[4] Marshall, Alfred. Principles of Economics.
London: Macmillan, 1890. Édition numérique disponible via EconLib.
https://www.econlib.org/library/Marshall/marP.html

 

[5] Abdul Latif Jameel Poverty Action Lab (J-PAL). Methodology: Randomized Evaluations.
Massachusetts Institute of Technology (MIT).
https://www.povertyactionlab.org/methodology

 

[6] Singer, Peter. “Famine, Affluence, and Morality.” Philosophy & Public Affairs, Vol. 1, No. 3 (1972), pp. 229–243.
Stable URL (JSTOR):
https://www.jstor.org/stable/2265052

 

[7] Kahneman, Daniel & Tversky, Amos. “Judgment under Uncertainty: Heuristics and Biases.” Science, Vol. 185, No. 4157 (1974), pp. 1124–1131.
https://science.sciencemag.org/content/185/4157/1124

 

[8] Darley, John M. & Latané, Bibb. “Bystander Intervention in Emergencies: Diffusion of Responsibility.” Journal of Personality and Social Psychology, Vol. 8, No. 4 (1968), pp. 377–383.
https://psycnet.apa.org/record/1968-08862-001

 

[9] Small, Deborah A.; Loewenstein, George; Slovic, Paul. “Sympathy and Callousness: The Impact of Deliberative Thought on Donations to Identifiable and Statistical Victims.” Psychological Science, Vol. 18, No. 2 (2007), pp. 143–148.
https://journals.sagepub.com/doi/10.1111/j.1467-9280.2007.01879.x

 

[10] GiveWell. How We Work – Cost-Effectiveness Analysis Methodology.
GiveWell Research, méthodologie en ligne.
https://www.givewell.org/how-we-work/our-criteria/cost-effectiveness

 

[11] Pryce, J.; Richardson, M.; Lengeler, C. “Insecticide-treated nets for preventing malaria.” Cochrane Database of Systematic Reviews (2018).
https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD000363.pub3/full

Vous aimeriez peut-être aussi lire...

Valeur d’une vie en France
Romain Barbe

What is the cost of a human life?

Human life is precious. It is natural to want to mobilise all our resources to save a life, even if it only prolongs a life by a week. But what happens when other people are also in danger, and our resources are not enough to help them all? As a society, we face practical limits that force us to make difficult decisions.

Lire l'article »
Illustration : évaluation incertaine de l'impact climatique des dons, entre modélisation et transparence sur le coût par tonne de CO2
Ombline Planes

Elementor #34549

Assessing the climate impact of donations: between modelling and uncertainty At Mieux Donner, we sometimes use a calculation that is strikingly simple: “€1 = 1

Lire l'article »